Hi all,
I had a question about DESeq2, or a little sanity check.
In my case I'm only interested in comparing how expression differs between conditions at specific time-points. I believe I can do comparisons with combined factors e.g. Responder_Week1 v NonResponder_Week1 - but I can also do LRT for full time-course comparisons? Is there a particular method that would be better suited? I believe that I can get similar results from LRT by using the names - however I think I've confused myself further - so I could use a hand.
#Code I've been using
#Using response to treatment and time-point as the design for DESeq
dds <- DESeqDataSetFromMatrix(countData = countdata, colData = coldata, design = ~response+ time + response:time)
#Using the time-point as the reducement, which means we are testing the effect of the response?
dds <- DESeq(dds, test = "LRT", reduced = ~time)
#Get the results,
results(dds) # However, this will be for all, but we can get a specific comparison by using name and Wald test
results(dds, name = "responseResponder.timeW1", test = "Wald")
#But then, is this comparing a Non-Responder to Responder at Week 1 or Responder at Week 1 against the Base time-point?
#OR, by combined factors, where condition = e.g. Responder_Week1
dds <- DESeqDataSetFromMatrix(countData = countdata, colData = coldata, design = ~condition)
#And now results is simply
results(dds, c("condition","Responder_Week1", "NonResponder_Week1"))
Please do correct me, direct me, or criticise me - I'm only hoping to learn more and do this correctly :)
I hope this makes sense, if there's any confusion please feel free to ask (I am confused myself, hence my reaching out!)
Thanks for the reply! I agree it's much easier to follow! I was wondering if it's "technically" the right way to do so, or if the values for significance testing will be widely skewed by one method or the other?
I've been testing with both methods for a particular dataset I'm interested in, and they seem to come out somewhat similar through both methods, although with some differences.